Should I use Zotero or Mendeley for a PhD?

The decision matters less than the decision to use one at all, and it matters most in the long run. A doctoral library accumulates for years, so the property to weigh is what happens to your data if the tool changes, is acquired, or shuts down.

On that axis Zotero has the stronger position. It is developed by a non-profit, the code is open, the local database is inspectable, and export is complete rather than partial. If it disappeared tomorrow, your library would still be on your disk in a readable form.

Mendeley is a competent tool owned by a large academic publisher, which is a genuine consideration for some researchers and a non-issue for others. It is worth knowing that the ownership question is why many departments have shifted their recommendation over the last decade, rather than any single missing feature.

On features they are close enough that the difference rarely decides anything. Both handle citation styles, both integrate with word processors, both sync. Zotero's extensibility through plugins is the practical advantage most people notice, since it means the tool adapts to unusual workflows rather than the reverse.

The numbers behind that closeness are worth seeing. A 2026 comparative review puts Zotero's free cloud allowance at 300 MB against Mendeley's 2 GB, and counts roughly 1,000 citation styles for Zotero against roughly 2,000 for Mendeley (Jin et al., 2026, p. 3). Mendeley wins both of those counts. Zotero wins on what you can change yourself, since the same review credits it with community plugins, advanced LaTeX integration, and direct edits to Citation Style Language files (Jin et al., 2026, p. 4). Storage and style counts are fixed limits, plugin support is not, which is why the second column matters more across a doctorate.

Each tool has a specific weakness rather than a general one. For Zotero, metadata captured from poorly structured webpages comes in incomplete or inconsistent and needs manual correction, and the review says Zotero lacks organizational depth once a library gets very large and complex (Jin et al., 2026, p. 3). For Mendeley, the collaboration features were reduced, because Elsevier discontinued several collaborative features, including Mendeley Web Groups, in 2021, and the review records academic concern about data ownership and long-term direction after the acquisition (Jin et al., 2026, p. 3). The same review declines to name a universal winner and says the choice tracks career stage, disciplinary norms, and institutional support (Jin et al., 2026, pp. 3-4).

Choose Mendeley if your day is spent reading and annotating PDFs in one window, since that is the workflow the review describes it handling best, covering import, organization, highlighting, commenting, and tagging across desktop, web, and mobile apps (Jin et al., 2026, p. 3). Choose Zotero if your day is spent collecting across journal sites, catalogs, and repositories, since the review reports one-click metadata capture with frequent automatic PDF retrieval, plus group libraries for cross-institutional work (Jin et al., 2026, p. 3). If your supervisor or lab already runs one of them, that settles it faster than any feature table.

Neither tool checks whether a reference is real. The review flags the absence of integrated verification of reference authenticity in both, and says this matters more now that AI-assisted writing can introduce fabricated or incorrect references into a draft (Jin et al., 2026, pp. 3-4). Your reference manager will format a hallucinated citation as cleanly as a genuine one, so verification stays your job, ideally by resolving the DOI before the entry enters the library.

The one thing worth doing regardless of choice is a periodic export. A monthly BibTeX or RIS dump into your backups costs nothing and turns any future migration into an import rather than a recovery.

How do I migrate a library from Mendeley to Zotero?

Export the library as BibTeX or RIS with the attached files, then import it. Expect metadata to need cleaning afterwards, particularly capitalization, author name formats, and journal titles. Migrate once, check a sample of fifty entries, and do not run both libraries in parallel.

Check the attachments specifically. File links are the part most likely to break, and discovering it three months later means reattaching PDFs by hand.

Do the migration at a natural break rather than mid-chapter, and keep the old library intact until you have written something substantial in the new one and confirmed the citations render correctly.

How do I stop duplicate references piling up?

Add references from one place only, run the duplicate detection monthly rather than yearly, and prefer a DOI lookup over importing from a search result page. Most duplicates come from adding the same paper from two different databases with slightly different metadata.

The duplicates that survive automatic detection are the ones where a field differs enough to look like a distinct work: a preprint and the published version, a chapter and the book, an article with and without a subtitle. Those need a manual scan, which is worth doing once a year.

Merging rather than deleting is the safer operation, since it preserves attachments, notes, and existing links to the entry from your writing.